Essential function of HIPK2 in TGFβ-dependent survival of midbrain dopamine neurons

Essential function of HIPK2 in TGFb-dependentsurvival of midbrain dopamine neurons

Jiasheng Zhang1,5,6, Vanee Pho1,5,6, Stephen J Bonasera2, Jed Holtzman3, Amy T Tang1,5, Joanna Hellmuth4,Siuwah Tang1,5, Patricia H Janak4, Laurence H Tecott3 & Eric J Huang1,5

Transforming growth factor beta (TGFb) is a potent trophic factor for midbrain dopamine (DA) neurons, but its in vivo function and

signaling mechanisms are not entirely understood. We show that the transcriptional cofactor homeodomain interacting protein

kinase 2 (HIPK2) is required for the TGFb-mediated survival of mouse DA neurons. The targeted deletion of Hipk2 has no

deleterious effect on the neurogenesis of DA neurons, but leads to a selective loss of these neurons that is due to increased

apoptosis during programmed cell death. As a consequence, Hipk2–/– mutants show an array of psychomotor abnormalities.

The function of HIPK2 depends on its interaction with receptor-regulated Smads to activate TGFb target genes. In support of

this notion, DA neurons from Hipk2–/– mutants fail to survive in the presence of TGFb3 and Tgfb3–/– mutants show DA neuron

abnormalities similar to those seen in Hipk2–/– mutants. These data underscore the importance of the TGFb-Smad-HIPK2 pathway

in the survival of DA neurons and its potential as a therapeutic target for promoting DA neuron survival during neurodegeneration.

Ventral midbrain DA neurons in the substantia nigra pars compacta(SNpc) and ventral tegmental area (VTA) control extrapyramidalmovement and a variety of cognitive functions, such as motivation,drive, addiction and reward behaviors. The selective loss of DAneurons, or perturbations to the neural circuits established by DAneurons, such as the nigrostriatal and mesolimbic pathways, have beenimplicated in several neurodegenerative conditions, including Parkin-son’s disease1,2. After the initial phase of neurogenesis, DA neuronsundergo a period of programmed cell death (PCD) during the perinatalstages. In rodents, the majority of PCD in DA neurons occurs duringthe perinatal stages, from postnatal day 0 (P0) to P4, with a smalleramount of PCD detected at P14 (ref. 3). Despite the occurrence ofPCD, the number of neurons in the SNpc, as detected by Nissl ortyrosine hydroxylase (TH) immunoreactivity, progressively increasefrom P0 to P7 and reaches a plateau at P14.

Although the detailed mechanisms that regulate PCD in DA neuronsare unknown, it has been proposed that, in a manner similar to sensoryneurons and spinal motoneurons4, neurotrophic factors and theirsignaling mechanisms may promote the survival of DA neurons.Indeed, several pro-survival trophic factors for DA neurons havebeen identified by using in vitro cultures of dissociated DA neuronsfrom the ventral midbrain of mouse embryos at embryonic days 13.5–14.5 (E13.5–14.5). These include brain-derived neurotrophic factor5,neurotrophin-4 (ref. 6), glial cell line–derived neurotrophic factor(GDNF)7 and TGFb8,9. Most of these trophic factors have well-documented functions in regulating the survival of sensory, sympa-thetic and parasympathetic neurons during neurogenesis10,11, which

raises the possibility that they may also provide survival cues to DAneurons during development4,12 and thereby control the final numberof DA neurons.

Several recent studies provide supporting evidence that links neuro-trophic factors to PCD in DA neurons. For instance, expression of theGDNF receptors c-ret and glial cell line–derived neurotrophic factorfamily receptor a1 (GFRa1) has been detected in midbrain DA neuronsduring development and in postnatal brains. As would be expectedfrom a target-derived trophic factor, GDNF mRNA is detected in themedium-sized striatal neurons, the presumptive target of DA neurons,at perinatal stages, but its concentration is significantly reduced afterP14. The ectopic expression of GDNF in the striatum either by directinjection or by a transgenic approach reduces the magnitude of PCD inthe DA neurons of the SNpc at perinatal stages13,14. Conversely, theinjection of a neutralizing antibody to GDNF enhances cell deathduring the first wave of PCD, but has no detectable effect on the secondwave13. Although these results support the notion that GDNF is a pro-survival trophic factor, experiments with mice that have a targeteddeletion in Gdnf or Gfra1 show no evidence of a neuronal deficiency inthe SNpc at P015–17. The exact cause for this discrepancy is still unclear.It is possible that the early perinatal lethality of Gdnf and Gfra1 mutantsmay have prevented the examination of the effect of GDNF in postnatalDA neurons. Alternatively, additional trophic factors could act inde-pendently of, or synergistically with, GDNF to promote the survival ofmidbrain DA neurons during PCD.

One potential candidate for a pro-survival trophic factor is TGFb,which promotes the survival of cultured DA neurons with a higher

Received 4 October; accepted 17 November; published online 10 December 2006; corrected after print 25 January 2007; doi:10.1038/nn1816

1Department of Pathology, 2Medicine, 3Psychiatry, and 4Neurology, University of California San Francisco, 505 Parnassus Avenue, San Francisco, CA 94143, USA.5Pathology Service 113B, VA Medical Center, 4150 Clement Street, San Francisco, CA 94121, USA. 6These authors contributed equally to this work. Correspondenceshould be addressed to E.J.H. ([email protected]).

NATURE NEUROSCIENCE VOLUME 10 [ NUMBER 1 [ JANUARY 2007 77

ART ICLES©

2007

Nat

ure

Pub

lishi

ng G

roup

ht

tp://

ww

w.n

atur

e.co

m/n

atur

eneu

rosc

ienc

e

potency than GDNF8,9,18. Like GDNF, at least two TGFb iso-forms, TGFb2 and TGFb3, are expressed in the target tissues or inthe vicinity of the midbrain DA neurons during development19,20.Furthermore, application of a TGFb blocking antibody to chickembryos leads to a severe loss of DA neurons during development8.Despite these results, the exact role of TGFb in the developmentof DA neurons remains poorly characterized. Studies from non-neuronal cells have indicated that the diverse functions of TGFb largelydepend upon the formation of multimeric protein complexes betweendifferent subtypes of Smad and their interacting partners, which in turnregulate the transcription of TGFb target genes21. HIPK2 is a tran-scriptional cofactor that has been recently shown to directly interactwith receptor-regulated Smads (R-Smads), including Smad1, Smad2and Smad3, and that enhances the suppressor functions of Smad1 onactivation by bone morphogenetic protein (BMP)22. As Smad2 andSmad3 are primarily activated by TGFb, however, it is possible thattheir interaction with HIPK2 may be required for signaling by theTGFb pathway.

In this study, we provide evidence that TGFb3 and HIPK2regulate the survival of DA neurons during PCD. The loss ofeither TGFb3 or HIPK2 results in similar deficiencies in the DAneurons, which is due to an increase in apoptosis. As a consequence,Hipk2–/– mutants show a severe psychomotor behavioral phenotypethat reflects neuronal deficiencies in the SNpc and VTA. HIPK2’sfunction depends on its ability to interact with and activate thedownstream targets of Smad3. Deletion of the Smad3-interactingdomain in HIPK2 completely abolishes the ability of HIPK2 to activateSmad3-dependent gene expression. Taken together, the resultsfrom this study provide the first evidence that the TGFb-HIPK2signaling mechanism regulates the survival and apoptosis of DAneurons during PCD.

RESULTS

Selective loss of midbrain DA neurons in

Hipk2–/– mutants

Using a reporter-gene construct, HIPK2lacZ

(ref. 23), which contains the b-galactosidasegene (lacZ), targeted in-frame with the firstcoding exon of Hipk2, we found that thisconstruct was expressed in the SNpc (A9)and VTA (A10), regions enriched with DAneurons (Fig. 1a–f). The expression ofHIPK2lacZ was detected in these two regions

as early as E15.5, became more prominent at E17.5 and remained highthroughout all of the postnatal stages (Fig. 1d–f). Confocal micro-scopic analyses showed that essentially all of the DA neurons (markedby TH) in these two areas coexpressed HIPK2lacZ (Fig. 1g–l). Incontrast to its expression in the SNpc and VTA, HIPK2lacZ was notdetected in the locus ceruleus or in target tissues that are innervated bymidbrain DA neurons, such as the striatum, globus pallidus or nucleusaccumbens (Supplementary Fig. 1 online). Unlike in sensory neu-rons23, expression of HIPK2 in the DA neurons of the SNpc and VTAdid not colocalize with POU domain, class 4, transcription factor 1(Pou4f1), also known as Brn3a. In fact, Brn3a was detected in a domaindorsolateral to the developing SNpc and VTA at E12.5 and in the rednucleus and interpeduncular nucleus at E15.5 and E17.5 (Fig. 1m–o).Taken together, these results indicate that DA neurons in the SNpc andVTA provide a system with which to investigate HIPK2 functions thatare independent of Brn3a.

To determine whether the loss of HIPK2 affected the development ofmidbrain DA neurons, we analyzed the cellular and volumetric contentsof the SNpc and VTA by stereology. Consistent with previous reports3,we found that the number of DA neurons in the SNpc and VTA in wild-type mice increased progressively during embryogenesis, reaching itsmaximum at P14, with a slight decrease in older mice (Fig. 2). Althoughthe number of midbrain DA neurons in wild type and Hipk2–/– mutantsseemed to be similar at E12.5, a substantial reduction was detectedin Hipk2–/– mutants as early as P0 and persisted into adulthood(Fig. 2b,d,g,h). Specifically, a 40% loss of DA neurons was detected inthe SNpc of Hipk2–/– mutants throughout the postnatal stages, whereasthe VTA showed a more pronounced deficit at P14, followed by a slightincrease in the number of DA neurons at P28 (Fig. 2g,h). The loss of DAneurons in Hipk2–/– mutants seemed to involve the entire SNpc andVTA (Fig. 2a,c and b,d, respectively). Consistent with the substantial

aF

FF

F

FF

SNpc

SNpc

SNpc

E12.5

Brn3a TH

IIIV

VTA

SNpc SNpcVTA

IPN

RNRN

E15.5 E17.5

TH THMerge Mergeβ-Gal β-Gal

VTA

VTA

E17.5

P28

Nissl HIPK2LacZ TH

b c

d

g h i j k l

e f

m n o

Figure 1 Expression of HIPK2 in midbrain DA

neurons during development and in postnatal life.

(a–f) HIPK2LacZ can be detected in the midbrain

DA neurons at E17.5, as seen in b, and at P28, as

seen in e. Adjacent brain sections are stained with

Nissl or antibody to TH to highlight the location of

the SNpc and VTA. The P28 brain section shows

the level at Bregma –2.9. F indicates fornix. Scalebar, 250 mm. (g–l) Confocal images showing

colocalization of TH and b-galactosidase in the

SNpc, seen in panels g–i, and VTA, seen in panels

j–l, at P28. Scale bar, 20 mm. (m–o) Immuno-

fluorescence staining in the ventral mesence-

phalon of an E12.5 embryo shows that Brn3a

expression is present adjacent to, but not

overlapping with, DA neurons, as seen in panel m.

As development progresses, Brn3a expression

becomes more restricted to the red nucleus (RN)

and the interpeduncular nucleus (IPN), but not in

the SNpc or VTA. Scale bar, 250 mm.

78 VOLUME 10 [ NUMBER 1 [ JANUARY 2007 NATURE NEUROSCIENCE

ART ICLES©

2007

Nat

ure

Pub

lishi

ng G

roup

ht

tp://

ww

w.n

atur

e.co

m/n

atur

eneu

rosc

ienc

e

loss of neurons, we observed a reduction of 43% and 32% in thevolumes of the SNpc and VTA of the Hipk2–/– mutants, respectively, anda 20% reduction in striatal dopamine content, by HPLC (Fig. 2i anddata not shown). The calbindin-D28K–expressing DA neurons, whichare relatively spared in Parkinson’s disease24, were also reduced, by 57%and 42% in the SNpc and VTA of Hipk2–/– mutants at P28, respectively(Fig. 2e,f, j). In contrast to the DA neuron deficiencies seen in the SNpcand VTA, there was no detectable loss of GABAergic neurons in thesubstantia nigra pars reticularis (Supplementary Fig. 2 online).

Psychomotor behavioral abnormalities in Hipk2–/– mutants

Perturbations of dopamine transmission resulting from deletion of theTh gene, blockade of dopamine re-uptake or neurotoxin treatment havebeen shown to produce psychomotor behavioral abnormalities25,26.Given the substantial loss of DA neurons in Hipk2–/– mutants, we per-formed a battery of tests to evaluate behavioral correlates of nigrostriatalfunction. We found that the majority of Hipk2–/– mice showed dystonia,characterized by the clasping of hindlimbs when mice were suspendedby their tails (Fig. 3a,b). In a 2-min test, 485% of Hipk2–/– mutantsshowed significantly longer durations of clasping compared to wild-typemice (Fig. 3c; P¼ 0.001). In addition, Hipk2–/– mutants had a shufflinggait, characterized by a reduced gait width and a consistent failure tofinish the tandem walk compared with wild-type mice (Fig. 3d). Thedistance between the hind paws was substantially reduced in Hipk2–/–

mutants, similar to that observed in MPTP-induced neurotoxicity27. Weobserved that Hipk2–/– mutants showed poor motor coordination on

the Rotorod test (Fig. 3e). Furthermore, Hipk2–/– mutants showedsevere retardation in the initiation of voluntary movement in four-limbakinesia and negative geotaxis tests (Fig. 3f,g).

In addition to the dystonia and impaired coordination, Hipk2–/–

mutants showed reduced responses to novelty compared with wild-type mice. In contrast to the active exploratory behaviors of wild-typemice, Hipk2–/– mutants showed reduced locomotion when placed in anunfamiliar environment. To quantify the difference, we used an open-field apparatus to measure novelty-related locomotion in wild type andHipk2–/– mutants. Compared with controls, Hipk2–/– mutants traveleda smaller total distance, entered the center of the field less frequentlyand spent more time in the periphery (Fig. 4a,b). As a result, thedistance of travel by Hipk2–/– mutants was markedly reduced in thecentral portion of the test field, but only modestly affected in the peri-pheral areas (Fig. 4c). The total distance traveled was significantlyreduced in Hipk2–/– mutants (P ¼ 0.003), whereas their average travelvelocity was only modestly reduced (Fig. 4d). After 5 d of acclimation,however, wild type and Hipk2–/– mutants showed no detectabledifference in baseline motor activity (data not shown). Notably,treatment with the dopamine type 1 (D1) receptor agonist SKF81297 increased similar home cage activity in wild-type and Hipk2–/–

mice, whereas injection with saline caused no difference in motoractivity (Fig. 4e). These results suggested that the function of thepostsynaptic D1 receptor remained intact in the absence of HIPK2. Todetermine whether the loss of mesencephalic DA neurons and thereduction of DA in the striatum of Hipk2–/– mutants affect responses

Wild type Wild type (n = 3)

VTAVTA

SNpca b

c d

g

h

i je f

VTA

VTASNpc

**

*

*

VTASNpc

4,000

3,000

Tota

l TH

+ n

euro

nsTo

tal T

H+ n

euro

ns

Vol

ume

(× 1

06 µm

3 )

Num

ber

of c

albi

ndin

+ c

ells

2,000

1,000

0

1,000

2,000

3,000

4,000

0

50

100

150

200

250

300

0

500

1,000

1,500

2,000

2,500

0

E12.5 P0 P14 P28 10 Months

E12.5 P0 P14 P28 10 Months

SNpcSNpc

TH Bregma –3.2

TH Bregma –3.6

Calbindin-D28K 50 µm 250 µm

Hipk2 –/–

HIPK2 –/– (n = 3)

Wild type

Hipk2 –/–

Wild type (n = 3)HIPK2 –/– (n = 3)

Figure 2 Loss of DA neurons in the SNpc and VTA of Hipk2–/– mutants. (a–d) Both the SNpc and VTA show a substantial reduction in the density of DA

neurons at P28. The deficiency appears to be quite uniform throughout the SNpc and VTA and shows no regional predilection. Images in panels a and b were

taken at the level of Bregma –3.2, whereas those in panels c and d were taken at Bregma –3.6. (e,f) The calbindin-D28K–expressing neurons in the SNpc and

VTA also show substantial reductions in Hipk2–/– mutants at P28. Scale bar: 250 mm. (g,h) The DA neuron deficiencies in the SNpc and VTA of Hipk2–/–

mutants are determined using unbiased stereology methods during development and in postnatal brains. The counts of the DA neurons at E12.5 represent theentire pool of TH+ neurons in the ventral mesencephalon. (i) The volumes of the SNpc and VTA are reduced in Hipk2–/– mutants. (j) The number of calbindin-

D28K–positive neurons shows a 57% and 43% reduction in the SNpc and VTA, respectively. Student’s t-test, n ¼ 3 for each genotype. * indicates P ¼ 0.005

for SNpc and 0.006 for VTA.


ART ICLES©

2007

Nat

ure

Pub

lishi

ng G

roup

ht

tp://

ww

w.n

atur

e.co

m/n

atur

eneu

rosc

ienc

e

to psychomimetic stimulants, we injected wild-type mice and Hipk2–/–

mutants with amphetamine, which inhibits DA reuptake and inducesDA efflux at nerve terminals. Locomotor activity was monitored in theopen field apparatus for 150 min. Hipk2–/– mutants showed a sub-stantially diminished response to amphetamine, whereas wild-typemice showed a marked increase in locomotor activity (Fig. 4f).

Loss of HIPK2 increases apoptosis in DA neurons during PCD

The timing and magnitude of DA neuron loss in the SNpc and VTA ofHipk2–/– mutants suggest that such deficits could result from either

impaired neurogenesis or increased cell death during development(Fig. 2g,h). The absence of any detectable difference in the number ofDA neurons in the SNpc or VTA ofHipk2–/– mutants at E12.5, however,suggests that loss of HIPK2 most likely did not affect the generation ofDA neurons during the early stages (Fig. 2g,h). To further examine theeffect of HIPK2 loss, we designed neuronal birth-dating experiments todetermine whether the initial stages in the neurogenesis of DA neuronswere affected in Hipk2–/– mutants. In the first experimental scheme, wepulse-labeled pregnant dams with BrdU at E10.5, collected the embryosat E11.5, E13.5 and E15.5, and counted the numbers of DA neurons

a

e f g

b c d

Wild type Wild type(n = 5) Wild type

Hipk2–/–

(n = 16)

Wild type(n = 7)

Wild type(n = 4)

Wild type(n = 6)

Hipk2–/–

(n = 5)Hipk2–/–

(n = 9)Hipk2–/–

(n = 8)

Hipk2–/– 1 Hipk2–/– 2

Hipk2–/–

70

Dur

atio

n of

cla

sp (

s)

Dire

ctio

n of

trav

el

60

50

40

*

*

**

30

20

10

Tim

e on

rot

orod

(s)

Tim

e in

four

-lim

b ak

ines

ia te

st (

s)

Tim

e in

neg

ativ

e ge

otax

is te

st (

s)

20

40

60

80

100

120

140

160

180

10

100

200

300

400

500

600

0

100

150

200

250

50

0

a b c d

e f

Wild type Hipk2–/–

Central

600

500

400

300

*

*

**

* **

*

**

*

*

**

200

100

0

3,000

2,500

2,000

1,500

1,000

500

0

1,000

1,500

2,000

2,500

3,000

3,500

500

0

1,000

2,000

3,000

4,000

5,000

6,000

7,000

0

0 20 40 60 80 100Time (min)

Salineinjection

SKF 81297 injection

10 30 50 70 90 110 130 150

Amphetamineinjection

Tota

l dis

tanc

e tr

avel

ed (

cm)

Tota

l dis

tanc

e tr

avel

ed (

cm)

Num

ber

of b

eam

bre

aks

120 140 160 180

0

1

2

3

4

Trav

el v

eloc

ity(c

m s

–1)

200

400

600

800

1,000

1,200

0Peripheral

Tim

e sp

ent (

min

)

Dis

tanc

e tr

avel

ed (

cm)

Central Peripheral

Wild type (n = 7)

Hipk2 –/– (n = 5)

Wild type(n = 8)Hipk2 –/–

(n = 6)

Wild typeHipk2 –/–




Figure 4 Reduced response to unfamiliar environment and to amphetamine

challenge in Hipk2–/– mutants. (a) Although wild-type mice covered the entire

territory in open-field test cages, Hipk2–/– mutants tended to stay in corners

and rarely crossed the center. Representative locomotive activity paths from

two wild-type mice and two Hipk2–/– mutants are presented. (b,c) Compared

with wild type, Hipk2–/– mutants spent less time in the central portion of the

field and more time in the periphery (b). The distance traveled in the central

portion of the open field is reduced in Hipk2–/– mutants (c). Multivariant

ANOVA test, n ¼ 7 for wild type and n ¼ 5 for Hipk2–/– mutants. *P ¼ 0.013;

**P ¼ 0.32. (d) The total distance traveled was reduced in Hipk2–/– mutantsin the first 3 h of recording (20,367.46 ± 6,557 cm in wild type versus

5,337.22 ± 830 cm in Hipk2–/– mutants). ANOVA test, P ¼ 0.003, n ¼ 7 for

wild type and n ¼ 5 for Hipk2–/– mutants. A trend for travel velocity to be reduced in Hipk2–/– mutants is depicted in the inset, P ¼ 0.06. (e) Wild type and

Hipk2–/– mutants showed no difference in motor activity after acclimation to test cages (data not shown). Injection with D1 agonist SKF 81297 increased

overall activity in wild-type (n ¼ 8) and Hipk2–/– (n ¼ 6) mice. Cumulative beam breaks over 90 min from initial injection of vehicle (left) and SKF 81297

(right) are shown on the y axis. * indicates that the total beam breaks after SKF 81297 administration are significantly different from vehicle controls by

two-way ANOVA (F1,24 ¼ 24.08, P o 0.001). (f) Compared to wild-type mice, Hipk2–/– mutants showed reduced motor activity after amphetamine injection.

Two-way repeated-measures ANOVA revealed a significant interaction of genotype versus time; F14, 196 ¼ 4.70, P o 0.0001; * indicates significant difference

at the given time points, P o 0.05, n ¼ 8 for each genotype.

Figure 3 Motor behavioral abnormalities in

Hipk2–/– mutants. (a–c) The majority of Hipk2–/–

mutants showed hindlimb clasping when

suspended by their tails. In a 2-min test, Hipk2–/–

mutants spent about 60 s in the clasping position.

Student’s t-test, P ¼ 0.001, n ¼ 16 for Hipk2–/–

mutants and n ¼ 5 for wild type. (d) Hipk2–/–

mutants also had gait abnormalities, including aninability to initiate tandem walk and a severe

hesitation during the test. (e) Rotorod test showed

that wild-type mice (n ¼ 7) performed significantly

better than Hipk2–/– mutants (n ¼ 5). Repeated-

measures ANOVA, * indicates P ¼ 0.001. (f,g) For

four-limb akinesia tests, as seen in panel f, mice

were placed on a brightly lit table and the amount

of time taken for the mice to move all four limbs

was recorded (n ¼ 4, wild type and n ¼ 9,

Hipk2–/–). In the negative geotaxis test, seen in g,

mice were placed head-down on a platform tilted

at 601 and the time for the mice to turn around

and move up toward the top was recorded (n ¼ 6,

wild type and n ¼ 8, Hipk2–/– mutants)25. Under

both tests, Hipk2–/– mutants performed signifi-

cantly worse than wild type mice. One-way ANOVA,

* indicates P ¼ 0.001, f, and P ¼ 0.038, g.


ART ICLES©

2007

Nat

ure

Pub

lishi

ng G

roup

ht

tp://

ww

w.n

atur

e.co

m/n

atur

eneu

rosc

ienc

e

(Fig. 5a). Our results indicated that, in wild-type mice, the percentagesof TH+BrdU+ neurons, which represents the nascent DA neuronsgenerated within the chase periods, were 82%, 89% and 76% atE11.5, E13.5 and E15.5, respectively (Fig. 5b,d,f,l). Notably, there wasno detectable difference between the wild type and Hipk2–/– mutants(Fig. 5b–g,l). We used a second protocol, with injections of BrdU atE15.5, to examine the later phase of neurogenesis. Under this condi-tion, however, no TH+BrdU+ neurons were detected at E16.5, E18.5 orP0 in the wild type or Hipk2–/– mutants (data not shown). The lack ofcolabeling may be due to a prolonged cell cycle in the progenitor cells,which may have exceeded the detection time frame. To further studythe role of HIPK2 in the neurogenesis and differentiation of DAneurons, we examined the expression of several DA neuron–specificmarkers, including nuclear receptor subfamily 4, group A, member 2(Nurr1), paired-like homeodomain transcription factor 3 (Pitx3),neurogenin 2 (Ngn2) and DA transporter (DAT), at various stagesduring development and in postnatal life. Our results showed noreduction in the expression of these markers in DA neurons ofHipk2–/– mutants (Fig. 5h–k and Supplementary Fig. 3 online).Using Pitx3 as an independent marker, we found that the number ofPitx3+ cells in the SNpc and VTA at P0 were reduced by 42% and 48%,

respectively, similar to what was determined using TH as marker(Fig. 5m and Fig. 2g,h).

To examine whether the loss of DA neurons in Hipk2–/– mutantsresults from an increase of cell death during development, we usedactivated caspase 3 as a marker to determine the extent of PCD inHipk2–/– mutants3. Consistent with previous results, we found thatapoptosis in wild-type mice reached its peak in both the SNpc and VTAat E17.5 and P0 (Fig. 6). Although the extent of apoptosis in DAneurons was similar between wild type and Hipk2–/– mutants at E12.5and E14.5, it was significantly higher (Po 0.05) in Hipk2–/– mutants atE17.5 and P0 (Fig. 6b,d,e, arrows). After P0, there was no detectable celldeath in the SNpc or VTA in the postnatal brains of wild-type mice andHipk2–/– mutants (Fig. 6e). Many of the apoptotic cells in Hipk2–/–

mutants seemed to lack immunoreactivity for TH at lower magnifica-tions, with only a few showing colabeling for activated caspase 3 andTH (Fig. 6b,d). To determine whether the lack of colabeling was due tothe different subcellular localization of activated caspase 3 and TH, weused a series of confocal images taken at incremental 2-mm intervals inthe z positions, which showed partial or extensive colabeling of caspase 3and TH in several apoptotic DA neurons of Hipk2–/– mutants(Fig. 6f,h). Similarly, DA neurons also showed coexpression of

a

b c

h i

j

l m

k

d e

f g

Embryonic day

3x BrdUinjections

10.5 11.5

Wild type Hipk2 –/–

Wild type Hipk2 –/–

Wild type (n = 3)

E11.5(n = 3)

0

20

0

500

1,000

1,500

**

2,000

40

60

80

100

Embryoscollected at:

E13.5(n = 2)

E15.5(n = 3)

SNpc VTA

Hipk2–/–

(n = 3)

Wild type

Hipk2–/–

E10.5 → E11.5

Ngn2

Pitx3

F FSNpcSNpc

VTA

BrdU injection at E10.5

Per

cent

age

Brd

U+T

H+ c

ells

Num

ber

of P

itx3+

cel

ls

VTA

Ngn2

E10.5 → E13.5

E10.5 → E15.5

BrdU TH

Ngn2 TH

12.5 13.5 14.5 15.5 16.5 17.5 18.5 19.5 20.5

Figure 5 Loss of HIPK2 does not affect neurogenesis in DA neurons. (a) Schematic diagram of neuronal birth-dating experiments. To examine the early

phases of neurogenesis, we pulse-labeled pregnant dams with 3 BrdU injections at E10.5 or E15.5 and embryos were collected 1, 3 or 5 d after injection.

(b–g) Double immunofluorescence for BrdU (red) and TH (green) shows no difference between the total number of DA neurons and the percentage of DA

neurons that have incorporated BrdU at E11.5, E13.5 and E15.5. Insets in panels d to g are higher magnifications of the enlarged areas. Arrowheads indicate

TH+BrdU+ neurons and arrows indicate TH+BrdU– neurons. Scale bar, 50 mm (b,c), 100 mm (d–g), 12.5 mm (insets in d–g. (h,i) Double immunofluorescence of

Ngn2 (red) and TH+ neurons (green) at E12.5 shows no detectable difference between wild type and Hipk2–/– mutants. Scale bar, 50 mm. (j,k) The expression

of Pitx3 shows no detectable reduction in the number of in Hipk2–/– DA neurons at P0, although the density of Pitx3+ cells is reduced. Scale bar, 250 mm.

(l) The percentage of TH+BrdU+ neurons shows no difference between wild type and Hipk2–/– mutants. (m) Quantification of Pitx3+ neurons shows a 42%

and 48% reduction in the SNpc (P ¼ 0.018) and VTA (P ¼ 0.002) of Hipk2–/– mutants at P0, respectively. Student’s t-test, n ¼ 3 for each genotype.


ART ICLES©

2007

Nat

ure

Pub

lishi

ng G

roup

ht

tp://

ww

w.n

atur

e.co

m/n

atur

eneu

rosc

ienc

e

caspase 3 with another DA neuron marker, Nurr1 (SupplementaryFig. 4 online). The increase of apoptosis in the SNpc and VTA ofHipk2–/– mutants at E17.5 and P0 coincided with the reduction of DAneurons at similar stages and supports the notion that the loss ofHIPK2 leads to an increase in cell death in midbrain DA neurons.

HIPK2 cooperates with R-Smads to regulate TGFb targets

The identification of HIPK2 as a direct interacting partner withreceptor-activated Smads led us to hypothesize that HIPK2 is anessential component of the TGFb-mediated signaling pathways thatsupport the survival of DA neurons. To test this hypothesis, we culturedDA neurons from the ventral mesencephalon of E13.5 wild-type andHipk2–/– mutant embryos. Our goal was to determine whether the lossof HIPK2 led to a compromise in the survival of neurons cultured incertain trophic factors. Similar to published results8,18, we found thatboth TGFb3 and FGF8 were robust survival signals for wild-type DA

neurons in vitro, whereas the effect of GDNFwas modest (Fig. 7a). Notably, although DAneurons from Hipk2–/– mutants survived inthe presence of FGF8 and GDNF, they failed tosurvive in the presence of TGFb3 (Fig. 7a). Tofurther investigate whether the loss of HIPK2led to selective defects in the responses todifferent subtypes of TGFb, we culturedwild-type and Hipk2–/– DA neurons withTGFb1, TGFb2 and TGFb3. We found thatTGFb3 was the most effective of the three atpromoting the survival of wild-type DA neu-rons, whereas TGFb1 and TGFb2 had a verymodest or no obvious effect, respectively(Fig. 7b). The presence of all three TGFbsdid not provide an additive or synergisticeffect for promoting the survival of thewild-type neurons, nor did they restore thesurvival of DA neurons from Hipk2–/–

mutants (Fig. 7b).Previous results have shown that HIPK2

directly interacts with R-Smads, includingSmad2 and Samd3, which raises the possibilitythat such interactions may be required forTGFb3-mediated survival in DA neurons22.To test this, we first determined the functionof HIPK2 in the downstream signaling path-ways of TGFb in HEK293 cells using a reporterconstruct, SBE-luc, that contains multipleSmad binding elements upstream of thefirefly luciferase gene28. We found thatHIPK2 synergized with R-Smad to activateSBE-luc. HIPK2-induced activation of SBE-luc was most pronounced in assays usingSmad3 alone or when Smad3 was used incombination with Smad2 or Smad4(Fig. 7c). This effect was further enhancedby TGFb3 (Fig. 7c).

The Smad-interacting domain in HIPK2has been mapped to the region betweenamino acids 635 and 867, which also con-tained interaction domains for homeodomainproteins, such as Brn3a and NK3 (refs.22,23,29). To identify the region(s) in HIPK2required to activate SBE, we performed a

series of deletions from the C-terminus of HIPK2. Our results showedthat the deletion of amino acids 1080–1189 (D1080–1189) or aminoacids 898–1189 (D898–1189) did not affect the ability of HIPK2 toactivate SBE. In contrast, these two mutants showed slightly higheractivities in SBE luciferase assays (Fig. 7d). These results were con-sistent with the findings that sequences between amino acids 976 and1189 can recruit transcriptional co-repressors, such as CtBP (G. Weiand E.J.H., unpublished data). Therefore, removing these sequencesincreased the coactivator effect of HIPK2. Further deletion of the C-terminal half of the Smad-interacting domain at amino acid 752(D752–1189) completely abolished the ability of HIPK2 to activateSBE (Fig. 7d). As the Brn3a-interacting domain (amino acids 752–897)in HIPK2 contains the C-terminal half of the Smad-interactingdomain, we reasoned that deletion of the Brn3a-interacting domainmight affect the interaction with Smad. Consistent with this prediction,deleting the Brn3a-interacting domain (DBID) completely abolished

a b e

c

f

g

h

dSNpc

VTA

0 µm +2 µm +4 µm +8 µm +10 µm

Merge

Caspase 3

TH

TH Caspase 3


Wild type (n = 4)

Hipk2–/– (n = 4)

100

75

50

25

0E12.5

Num

ber

of c

aspa

se 3

+ c

ells

*

*

E14.5 E17.5 P0 P14

Figure 6 Loss of HIPK2 leads to increased apoptosis of DA neurons during periods of PCD.

(a–d) Immunofluorescence microscopy shows an increase in apoptotic cell death, highlighted by the

activated caspase 3 antibody (red), in the SNpc and VTA of Hipk2–/– mutants. The presence of DA

neurons is marked by a TH antibody (in green). (e) Quantification of activated caspase 3–positive cells in

the SNpc and VTA highlights the window of PCD of DA neuron in wild-type mice at E17.5 and P0.

A significant increase of caspase 3–positive cells is noted in Hipk2–/– mutants at the peak of PCD.

Student’s t-test, P o 0.05, n ¼ 4 for wild type and Hipk2–/– mutants. (f–h) A series of confocal images,

taken at incremental 2-mm intervals in the z positions to demonstrate the colocalization of caspase 3 and

TH in 2 apoptotic neurons in the VTA of Hipk2–/– mutants (arrows). Although the neuron on the left

shows extensive colabeling of caspase 3 and TH, only partial colocalization is present in the neuron on

the right (arrowhead). Scale bar, 20 mm in h.


ART ICLES©

2007

Nat

ure

Pub

lishi

ng G

roup

ht

tp://

ww

w.n

atur

e.co

m/n

atur

eneu

rosc

ienc

e

the coactivator activity of HIPK2 (Fig. 7d). Notably, deleting thespeckle-retaining sequence also abolished the activity of HIPK2,probably by preventing HIPK2 localization to the nuclear body30. Incontrast, inactivation of the kinase activity of HIPK2 by altering thehighly conserved lysine residue in the ATP-binding pocket(HIPK2K221A) did not alter the effect of HIPK2, suggesting that thekinase activity of HIPK2 may not be necessary for its function as acoactivator (Fig. 7d).

DA neuron deficiencies in Tgfb3–/– mutants

The distinctive effect of TGFb3 on DA neuron survival suggests thatTGFb3 may also provide pro-survival signals to DA neurons duringPCD in vivo. Therefore, on the basis of the proposed function ofHIPK2, one would predict that the removal of TGFb3 should result in aphenotype similar to that observed in Hipk2–/– mutants. To test thishypothesis, we first examined the expression of TGFb1 and TGFb3 inthe ventral mesencephalon during the periods of PCD. Using isoform-specific antibodies, we found that TGFb3 was expressed at a highconcentration in regions that are adjacent to DA neurons (Fig. 8a–c).In contrast, immunoreactivity for TGFb1 was barely detectable(Fig. 8b). We then examined the extent of apoptosis and cell loss inthe midbrain DA neurons of mice lacking TGFb1 or TGFb3. Mice withthe targeted deletion of Tgfb3 were maintained in the same genetic

background as Hipk2–/– mutants (C57B6;129SvJ). Tgfb3–/– mutants undergo perinatallethality because of craniofacial defects31,32,limiting our analyses of DA neuron deficien-cies to prenatal and perinatal stages. Analysesof DA neurons showed that, similarly to whatoccurs in Hipk2–/– mutants, the loss of TGFb3did not affect the total number of midbrainDA neurons at E12.5, but it led to a 40–50%reduction in the number of DA neurons in theSNpc and VTA at P0 (Fig. 8d–f). The numberof DA neurons in the SNpc and VTA ofTgfb3–/– mutants was almost identical to thosein Hipk2–/– mutants at P0 (Fig. 8f andFig. 2g,h). Corresponding to the neuronaldeficiencies, a twofold increase in activatedcaspase 3–positive neurons was identified inthe SNpc and VTA of Tgfb3–/– mutants at thesame stages (Fig. 8g–i).

Unlike the Tgfb3–/– mutants, mice with atargeted deletion of Tgfb1 showed no grossabnormalities at birth and, in the NIH-Olacgenetic background, survived more than1 month after birth, but suffered from auto-immune disorders and neurodegeneration inthe cortical neurons33–35. The differences ingenetic background between the Tgfb1–/– andHipk2–/– mutants did not affect the number ofDA neurons (Supplementary Table 1 online).Notably, Tgfb1–/– mutants showed no detect-able reduction in the number of DA neuronsin the SNpc or VTA at 4 weeks of age (Fig. 8j).Taken together, these results are consistentwith the selective effect of TGFb isoforms insupporting the survival of DA neurons in vitro(Fig. 7). The marked similarity of the DA neu-ron deficiencies betweenTgfb3–/– andHipk2–/–

mutants support the notion that signalingmechanisms involving TGFb3 and the Smad-interacting transcriptionalcofactor HIPK2 are required for the survival of DA neurons.

DISCUSSION

In this study, we present evidence that TGFb3 and the transcriptionalcofactor HIPK2 are essential for the survival of midbrain DA neurons.Our results indicate that both Hipk2–/– and Tgfb3–/– mutants show asubstantial loss of DA neurons that is due to increased apoptosis duringPCD. The prominent psychomotor behavioral abnormalities inHipk2–/– mutants are consistent with those described in mouse mutantslacking TH in DA neurons25,26. Injection with amphetamine, whichinduces DA efflux from nerve terminals, produced markedly reducedlocomotor activity inHipk2–/– mutants (Fig. 4f). In contrast, treatmentwith the D1 receptor agonist SKF 81297 restored locomotor activity inHipk2–/– mutants (Fig. 4e). Taken together, these results support thenotion that the motor phenotypes in Hipk2–/– mutants are dueprimarily to deficiencies in the number of DA neurons, whereaspostsynaptic D1 receptor function remains intact.

Despite the similarity of DA neuron deficiencies between Hipk2–/–

and Ngn2–/– mutants36,37, the mechanisms by which HIPK2 andNgn2 regulate the development of DA neurons are quite different.In contrast to the severe impairment of neurogenesis observed inNgn2–/– mutants from E11.5 to E14.5, the loss of HIPK2 does not affect

1.5

1.0

0.5

0

1.5

2.0

1.0

0.5

0

100

200

300

400

500

600

700

800

0

50

100

150

200

250

300

3500

SBESmad2Smad3Smad4

+–––

+–+–

+++–

+–++

++++

Contro

l

TH

-pos

itive

neu

rons

(%

)

TH

-pos

itive

neu

rons

(%

)

Luci

fera

se a

ctiv

ity

Relative luciferase activity

GDNF

** * *

TGFβ3FGF8

Contro

l

TGFβ2

TGFβ1

TGFβ3

β1β2

β3

KD

HIPK-WT

HIPK2K221A

∆1080–1189

∆898–1189

∆752–1189

∆SRS

∆BID

0 192 520

221

*

587 614 860 897

897

897

967

860 967

967

1066 1189

1189

752

752

752

2B BID SRS YH

Wild typeHIPK2 –/–

Wild type Smad onlya

d

b cSmad + HIPK2Smad + HIPK2 + TGFβ3

Smad only+ HIPK2 const.+ HIPK2 const. + TGFβ3

HIPK2 –/–

Figure 7 HIPK2 is important in the TGFb-Smad signaling pathway. (a) DA neurons from the ventral

mesencephalon of E13.5 embryos are cultured in the defined medium. DA neurons lacking HIPK2 fail to

survive in the presence of TGFb3, but survive in the presence of FGF8 or GDNF. (b) Among the three

TGFb isoforms, TGFb3 shows the most potent effect in supporting cultured DA neurons, whereas TGFb1

shows a very modest effect and TGFb2 has no detectable effect. Statistical analyses for panels a and b

use Student’s t-test, n ¼ 4 for wild-type and n ¼ 5 for Hipk2–/– mutants. * indicates P o 0.05.

(c) HIPK2 enhances the ability of R-Smads, especially Smad3, to activate luciferase reporter SBE in

HEK293 cells. The coactivator effect of HIPK2 is further enhanced by TGFb3 (5 ng ml-1). (d) Deletion of

the Brn3a-interaction domain (DBID), speckle-retention domain or both (D752–1189) completely

abolishes the coactivator effect of HIPK2 in SBE luciferase assays in HEK293 cells. In contrast, loss of

the co-repressor domain in the C-terminus (D1080–1189 or D898–1189) leads to a slight increase in

the SBE activity. Point mutation of the conserved lysine residue in the ATP binding pocket, HIPK2K221A,

does not affect the activity of HIPK2 in SBE luciferase assays. Data represent mean ± s.e.m. from three

independent experiments.


ART ICLES©

2007

Nat

ure

Pub

lishi

ng G

roup

ht

tp://

ww

w.n

atur

e.co

m/n

atur

eneu

rosc

ienc

e

the number of DA neurons at E12.5 (Fig. 2). In addition, the expressionof several cell fate and differentiation markers for DA neurons,including Nurr1, Pitx3, Ngn2 and DA transporter (DAT), was notdetectably reduced inHipk2–/– mutants (Fig. 5h–k and SupplementaryFig. 3). Furthermore, detailed neuronal birth-dating experimentsalso did not show any difference in neurogenesis between wild typeand Hipk2–/– mutants during embryogenesis (Fig. 5a–g). BrdU injec-tions at E15.5 fail to label any BrdU+TH+ neurons at E16.5, E18.5and P0. These results are consistent with a previous report thatprogenitors in the SNpc have limited potential to differentiate intoDA neurons38. It is possible that these progenitors may have prolongedcell cycles after E15.5. Alternatively, the neurogenesis of DA neuronsafter E15.5 may involve a longer maturation process. Regardless of theexact mechanism, the loss of HIPK2 has no detectable effect on theneurogenesis of DA neurons. In contrast, the timing and the magnitudeof apoptosis in DA neurons of Hipk2–/– mutants during PCD, and theselective failure of Hipk2–/– DA neurons to survive in TGFb3, supportthe hypothesis that HIPK2 is an important component in the TGFbsignaling pathway that regulates the survival of DA neurons (Supple-mentary Fig. 5 online).

The manner by which TGFb controls neuronal survival or apoptosisis cell type dependent. For instance, TGFb promotes the survival ofmidbrain DA neurons in dissociated cultures, and blocking TGFbsignaling with neutralizing antibody in chick embryos8,9 or through thetargeted deletion of Tgfb3 or Hipk2 in mouse (Figs. 2, 5 and 7) leads tosimilar reductions in the number of DA neurons. Notably, thedeficiencies in the number of DA neurons of treated chick embryosoccur after the induction of DA neurons, suggesting that theseparticularly sensitive stages may correspond to the periods of PCDidentified in rodents3. In contrast to its effects on DA neurons, areduction of TGFb signals produces a robust reduction in neuronalapoptosis in sensory neurons during PCD39. Consistent with theseresults, the targeted deletion of Hipk2 results in reduced apoptosis andincreased neuron number in the trigeminal ganglion during PCD23.

The phenotypes observed in the sensory ganglion and midbrain DAneurons of Hipk2–/– mutants are consistent with the function of TGFbin these neurons and further support the notion that the signalingmechanisms involving TGFb and HIPK2 are cell type specific.

Two lines of evidence further indicate that HIPK2 is an essentialelement in the downstream signaling pathway of TGFb. First, DAneurons lacking HIPK2 fail to survive in the presence of TGFb, but notin the presence of other trophic factors (Fig. 7). Second, HIPK2 directlyinteracts with R-Smads22, including Smad2 and Smad3, suggesting thatHIPK2 can modulate the signaling pathways of TGFb. Finally, we showthat HIPK2 enhances the ability of Smad3 to activate the SBE luciferaseconstruct. The presence of TGFb further promotes the coactivatoreffect of HIPK2 (Fig. 7c). Conversely, deletion of the Smad-interactingdomain completely abolishes the ability of HIPK2 to promote theactivity of Smad3 (Fig. 7d). Taken together, these results suggest thatthe interactions between HIPK2 and R-Smads are essential for DAneurons to respond to the survival cues of TGFb. Although it istempting to propose that the interactions between HIPK2 andR-Smads in the ‘canonical’ TGFb signaling pathway are essential forthe pro-survival function of TGFb (Fig. 6e), HIPK2 may regulate thesurvival of DA neurons through Smad-independent mechanisms. Forinstance, HIPK2 has been implicated in the p53-induced apoptoticpathway by the activation of p53 phosphorylation40,41. In addition,recent evidence has shown that the p53-mediated activation of caspase 6can further activate HIPK2 (ref. 42). Our results, however, indicate thatin mouse embryonic fibroblasts (MEFs), a more sensitive system for p53function, loss of HIPK2 has essentially no negative impact on p53phosphorylation (G. Wei and E.J.H., unpublished data).

Our results in Figure 7 and results from several previous studiesindicate that TGFb is a more potent pro-survival factor thanGDNF8,9,18. Among the three isoforms of TGFb, TGFb3 seems to bethe most efficacious in supporting DA neuron survival (Fig. 7). TGFb3is expressed at a relatively high concentration in regions close to DAneurons in the SNpc (Fig. 8) and therefore may provide local survival

VTASNpcVTASNpc

NS

NS4,000

Wild type (n = 3) Tgf�1–/– (n = 3)

Wild type

Wild typeTgf�3 –/–

Tgf�3 –/–

Wild type (n = 4)Tgf�3 –/– (n = 4)

Tgf�3 –/–

Tgf�3 –/–

3,000

2,000

1,000

0

Tota

l TH

+ n

euro

ns

80

70

60

50

40

30

20

10

0To

tal c

aspa

se 3

+ c

ells

P0 VTA(n = 4)(n = 3)

E12.5

P0 SNpc(n = 4)(n = 3)

E12.5

0

500

1,000

1,500

2,000

2,500

NS

NS

*

*

1,500

1,000

500

0

*

Tota

l TH

+ n

euro

nsTo

tal T

H+ n

euro

ns

THSNpc

SNpr

P0 ventral midbrain P0 ventral midbrain

TGFβ1 TGFβ3

200 µm 500 µm

25 µm

Wild type

Wild type

F F

SNpcSNpc

VTA VTA

TH Caspase 3

a f

i j

b c d e

g h

Figure 8 Loss of TGFb3, but not TGFb1, leads to DA neuron deficiencies similar to those in Hipk2–/– mutants. (a–c) Immunohistochemistry shows that TGFb1

is undetectable in the ventral mesencephalon at P0 (b). In contrast, a relatively high concentration of TGFb3(c) is present in regions adjacent to DA neurons in

the SNpc (a). (d,e) Reduction of DA neuron density is present in the SNpc and VTA of Tgfb3–/– mutants at P0. Scale bar, 500 mm. (f) Quantification of DAneurons in wild-type and Tgfb3–/– mutants shows no difference at E12.5, but does show a 50% reduction in both the SNpc and VTA at P0. * indicates two-

tailed Student’s t-test, P ¼ 0.018 for SNpc and P ¼ 0.002 for VTA, n ¼ 3 for E12.5 and n ¼ 4 for P0. (g,h) Increased apoptotic cell death is present in the

SNpc and VTA in Tgfb3–/– mutants at P0. Confocal images of double immunofluorescence using activated caspase 3 (red) and TH (green). Arrows indicate

double positive cells. Scale bar, 25 mm. (i) Quantification of DA neurons and activated caspase 3–positive cells in the SNpc and VTA of Tgfb3–/– mutants

shows deficits similar to those in Hipk2–/– at the same stage. (j) No detectable loss of DA neurons is found in the SNpc or VTA of Tgfb1–/– mutants at P28.

NS indicates not significant, two-tailed Student’s t-test, n ¼ 3 for each genotype.


ART ICLES©

2007

Nat

ure

Pub

lishi

ng G

roup

ht

tp://

ww

w.n

atur

e.co

m/n

atur

eneu

rosc

ienc

e

cues for these neurons at perinatal stages. In contrast, TGFb1 expres-sion is barely detectable in the ventral midbrain and its loss has nonegative impact on the survival of DA neurons (Fig. 8b,j). Theseresults, together with recent studies33, suggest that TGFb1 and TGFb3are required for the survival of cortical and DA neurons, respectively.Notably, a recent paper suggests that TGFb2 and TGFb3 are bothrequired for the proper development of DA neurons43. Thus, it remainsunclear as to why cultured DA neurons respond less favorably toTGFb1 and TGFb2. Several explanations could account for thisdiscrepancy. First, different TGFb isoforms may have distinct post-translational modifications, which may affect their efficacy and potencyin vitro. Second, isoform-specific receptor subtypes or co-receptorsmay be required for full activation by different TGFbs. The expressionof these receptors may be cell type specific or disrupted by the cultureconditions for DA neurons. Finally, TGFb isoforms may have stage-dependent effects on the proliferation, survival and differentiation ofDA neurons.

In addition to their pro-survival functions, TGFb signaling mechan-isms have also been implicated in the regulation of synaptic plasticity. Inthe synapses of Aplysia californica sensory-motor neurons, TGFb1 isnecessary and sufficient for the induction of long-term facilitation and,on acute application, activates MAP kinase and synapsin phosphoryla-tion44,45. Furthermore, genetic analyses in Drosophila melanogasterprovide direct evidence that mutations in BMP homolog gbb (glassbottom boat) and BMP receptors tkv (thickveins), sax (saxophone) andwit (wishful thinking) lead to a marked reduction in synaptic size, aber-rant synaptic ultrastructure and decreased synaptic transmission46–49. Inlight of these results, it is interesting to note that, although Hipk2–/–

mutants still retain about 50–60% of the DA neurons in the SNpc andVTA, they show marked motor deficits and reduced locomotorresponses to amphetamine (Fig. 3). These results suggest that TGFbsignaling mechanisms involving Smad and HIPK2 may also regulatesynaptic transmission in DA neurons.

In summary, we have identified HIPK2 as a transcriptional coacti-vator in the TGFb signaling pathway that regulates the survival ofmidbrain DA neurons during PCD. Our results indicate that interac-tion with Smad3 is essential for HIPK2 function and that the loss ofHIPK2 or TGFb3 results in similar robust neuronal deficits in ventralmidbrain DA neurons. These results suggest that the TGFb-HIPK2signaling pathway may serve as a potential therapeutic target forpromoting the survival of DA neurons.

METHODSHistology and immunohistochemistry. Detailed methods for histology,

immunohistochemistry and luciferase assays can be found in the Supplemen-

tary Methods online.

Stereology. The total numbers of TH+ neurons in the SNpc and VTA were

determined in wild type and Hipk2–/– mutants using the optical fractionator, an

unbiased cell counting method that is not affected by either the volume of

reference (for example, the SNpc or VTA) or the size of the counted elements

(for example, neurons). Neuronal counts were performed by using a computer-

assisted image analysis system consisting of an Olympus BX-51 microscope

equipped with a XYZ computer-controlled motorized stage and the Stereo-

Investigator software (Microbrightfield). TH-stained neurons were counted in

the left SNpc or VTA of every sixth section throughout the entire SNpc. Each

section was viewed at lower power (4�) and outlined. The numbers of

TH-stained cells were counted at high power (100�oil; NA 1.4) using a

50- � 50-mm counting frame. For P28 and P14 mice, a 15-mm dissector was

placed 2 mm below the surface of the section. For embryos and P0 mice, a

10-mm dissector was placed 1.5 mm below the surface of sections. Because

tissues at E12.5 were cut thinner, a fractionator method was employed in

counting TH+ neurons, where only cell profiles were counted, and is considered

as an estimate of cell population. The counts of DA neurons at E12.5

represented the entire pool of TH+ neurons in the ventral mesencephalon.

Measurements of locomotor activity. Mice (n ¼ 5 for Hipk2–/– mutants, n ¼7 wild type) were housed individually in photobeam activity cages (40 � 20 �20 cm, San Diego Instruments) for 48 hrs after they were allowed to habituate

in the same room for 7 d. Cages were equipped with 4 infrared photobeams,

spaced 8.8 cm apart along the long axis. For the open field test, mice were

placed in transparent Plexiglas cages for 10 min and locomotor activity was

monitored by video tracking software (Ethovision Tracking Software, Noldus

Information Technology) for 5 d. Details of the motor activity tests are

described in the Supplementary Methods.

Amphetamine challenge. Mice were habituated to the open field apparatus for

6 d until baseline locomotor activity showed no difference between the

genotypes. For each of these days, mice were injected with saline 30 min after

placement into the chamber and distance traveled over the ensuring 150 min

was recorded. On day 7, 30 min after placement into the chamber, mice were

injected with saline or 5 mg per kg (body weight) amphetamine and the

distance traveled was monitored for 150 min.

Locomotor response to SKF 81297. Mice were singly housed in low profile

cages and placed within standard photobeam frames. Mice had 4 d to

acclimatize to their home cages before testing began. On day 5, half of the

mouse cohort received an intraperitoneal injection of saline vehicle; the other

half received the D1 receptor agonist SKF 81297 (Sigma). The drug was

aliquoted to a dosage of 10 mg per kg, delivered in an injection volume of

10 ml g–1. After a further 4 d of acclimatization, all mice were tested a second

time. Mice that received vehicle injection on day 5 received SKF 81297 on day

11, and vice versa. No test session effect was observed. All injections occurred

between 15:00 and 16:00 on the testing day. Activity data (measured by

photobeam breaks) were collected for all mice for 90 min after all injections.

DA neuron cultures. Primary DA neurons were prepared according to

published procedures6,8. Briefly, E13.5 mouse embryos were collected from

time-pregnant Hipk2+/– females. The ventral mesencephalon was dissected,

dissociated after treatment with trypsin and cultured on cover slides coated with

poly-D-lysine and laminin. The dissociated cells were cultured in the presence of

a mixture of F12, DMEM, N1 additives (Sigma) and designated trophic factors.

In general, 2,500 cells were plated on each cover slide and allowed to grow in the

presence of GDNF (50 ng ml–1), TGFb (10 ng ml–1) or FGF8 (1 ng ml–1, R&D

Systems). Pilot cultures were established 2 hours after plating to determine the

total number TuJ1-positive cells and TH-positive cells. Representative results of

such cultures are presented (Supplementary Fig. 6 online).

Statistical analyses. Data were analyzed by two-tailed Student’s t-test (for

stereology counting (Figs. 2 and 7), clasping analyses (Fig. 3a), quantification

of apoptosis (Figs. 5 and –7) and DA neuron cultures and luciferase assays

(Fig. 6) and one-way or two-way ANOVA for behavioral tests (Figs. 3 and 4).

Values were expressed as mean ± s.e.m.

Note: Supplementary information is available on the Nature Neuroscience website.

ACKNOWLEDGMENTSWe thank L.F. Reichardt and S. Pleasure for comments on the manuscript.Special thanks to J. Siegenthaler for TGFb1 antibody, D.J. Anderson for Ngn2antibody, R. Akhurst for Tgfb1 mutants, T. Doestchman for Tgfb3 mutants,D. Di Monte for dopamine HPLC assays, T. Wyss-Coray for Smads and SBE-luciferase plasmids, A. Buckley for characterizing HIPK2LacZ expression, S. Ku formutant HIPK2 plasmids, C. Wong for histology and G. Wei for generating Hipk2mutants. V.P. was supported in part by a postdoctoral fellowship from theAmerican Parkinson Disease Association. This work was supported by grantsfrom the National Institute of Neurological Disorders and Stroke (NS44223 andNS48393), a Pilot Project Grant from the University of California San Francisco(UCSF) Alzheimer’s Disease Research Center (AG23501), Veteran’sAdministration Merit Review, Presidential Early Career Award for Scientists andEngineers, National Parkinson Foundation, and The Michael J. Fox Foundationfor Parkinson’s Research to E.J.H., funds from the state of California for medical


ART ICLES©

2007

Nat

ure

Pub

lishi

ng G

roup

ht

tp://

ww

w.n

atur

e.co

m/n

atur

eneu

rosc

ienc

e

research on alcohol and drug abuse through UCSF to P.H.J., and grants from theUS National Institutes of Health to P.H.J. and L.H.T.

AUTHOR CONTRIBUTIONSJ.Z., V.P., A.T.T. and S.T. performed the cellular, molecular and histologicalexperiments and the mouse genetics. L.H.T. supervised and S.J.B. and J.H.performed the Open Field, Rotorod and D1 agonist tests. P.H.J. supervisedand J.H. performed the amphetamine tests. J.Z., V.P. and E.J.H. wrotethe manuscript.

COMPETING INTEREST STATEMENTSThe authors declare that they have no competing financial interests.

Published online at http://www.nature.com/natureneuroscience

Reprints and permissions information is available online at http://npg.nature.com/

reprintsandpermissions/

1. Manji, H.K., Drevets, W.C. & Charney, D.S. The cellular neurobiology of depression.Nat.Med. 7, 541–547 (2001).

2. Dauer, W. & Przedborski, S. Parkinson’s disease: mechanisms and models. Neuron 39,889–909 (2003).

3. Jackson-Lewis, V. et al. Developmental cell death in dopaminergic neurons of thesubstantia nigra of mice. J. Comp. Neurol. 424, 476–488 (2000).

4. Oppenheim, R.W. Cell death during development of the nervous system. Annu. Rev.Neurosci. 14, 453–501 (1991).

5. Hyman, C. et al. Overlapping and distinct actions of the neurotrophins BDNF, NT-3, andNT-4/5 on cultured dopaminergic and GABAergic neurons of the ventral mesencepha-lon. J. Neurosci. 14, 335–347 (1994).

6. Hynes, M.A. et al. Neurotrophin-4/5 is a survival factor for embryonic midbraindopaminergic neurons in enriched cultures. J. Neurosci. Res. 37, 144–154 (1994).

7. Lin, L.F., Doherty, D.H., Lile, J.D., Bektesh, S. & Collins, F. GDNF: a glial cell line-derived neurotrophic factor for midbrain dopaminergic neurons. Science 260, 1130–1132 (1993).

8. Farkas, L.M., Dunker, N., Roussa, E., Unsicker, K. & Krieglstein, K. Transforming growthfactor-beta(s) are essential for the development of midbrain dopaminergic neurons invitro and in vivo. J. Neurosci. 23, 5178–5186 (2003).

9. Poulsen, K.T. et al. TGF beta 2 and TGF beta 3 are potent survival factors for midbraindopaminergic neurons. Neuron 13, 1245–1252 (1994).

10. Airaksinen, M.S. & Saarma, M. The GDNF family: signalling, biological functions andtherapeutic value. Nat. Rev. Neurosci. 3, 383–394 (2002).

11. Chao, M.V. Neurotrophins and their receptors: a convergence point for many signallingpathways. Nat. Rev. Neurosci. 4, 299–309 (2003).

12. Barde, Y.A. Trophic factors and neuronal survival. Neuron 2, 1525–1534 (1989).13. Oo, T.F., Kholodilov, N. & Burke, R.E. Regulation of natural cell death in dopaminergic

neurons of the substantia nigra by striatal glial cell line-derived neurotrophic factorin vivo. J. Neurosci. 23, 5141–5148 (2003).

14. Kholodilov, N. et al. Regulation of the development of mesencephalic dopaminergicsystems by the selective expression of glial cell line-derived neurotrophic factor in theirtargets. J. Neurosci. 24, 3136–3146 (2004).

15. Enomoto, H., Heuckeroth, R.O., Golden, J.P., Johnson, E.M. & Milbrandt, J. Develop-ment of cranial parasympathetic ganglia requires sequential actions of GDNF andneurturin. Development 127, 4877–4889 (2000).

16. Cacalano, G. et al. GFRalpha1 is an essential receptor component for GDNF in thedeveloping nervous system and kidney. Neuron 21, 53–62 (1998).

17. Enomoto, H. et al. GFR alpha1-deficient mice have deficits in the enteric nervoussystem and kidneys. Neuron 21, 317–324 (1998).

18. Krieglstein, K., Suter-Crazzolara, C., Fischer, W.H. & Unsicker, K. TGF-beta superfamilymembers promote survival of midbrain dopaminergic neurons and protect them againstMPP+ toxicity. EMBO J. 14, 736–742 (1995).

19. Flanders, K.C. et al. Localization and actions of transforming growth factor-beta s in theembryonic nervous system. Development 113, 183–191 (1991).

20. Unsicker, K., Flanders, K.C., Cissel, D.S., Lafyatis, R. & Sporn, M.B. Transforminggrowth factor beta isoforms in the adult rat central and peripheral nervous system.Neuroscience 44, 613–625 (1991).

21. Massague, J., Seoane, J. & Wotton, D. Smad transcription factors. Genes Dev. 19,2783–2810 (2005).

22. Harada, J. et al. Requirement of the co-repressor homeodomain-interacting proteinkinase 2 for ski-mediated inhibition of bone morphogenetic protein-induced transcrip-tional activation. J. Biol. Chem. 278, 38998–39005 (2003).

23. Wiggins, A.K. et al. Interaction of Brn3a and HIPK2 mediates transcriptional repressionof sensory neuron survival. J. Cell Biol. 167, 257–267 (2004).

24. Yamada, T., McGeer, P.L., Baimbridge, K.G. & McGeer, E.G. Relative sparing inParkinson’s disease of substantia nigra dopamine neurons containing calbindin-D28K. Brain Res. 526, 303–307 (1990).

25. Zhou, Q.Y. & Palmiter, R.D. Dopamine-deficient mice are severely hypoactive, adipsic,and aphagic. Cell 83, 1197–1209 (1995).

26. Szczypka, M.S. et al. Feeding behavior in dopamine-deficient mice. Proc. Natl. Acad.Sci. USA 96, 12138–12143 (1999).

27. Fernagut, P.O., Diguet, E., Labattu, B. & Tison, F. A simple method to measurestride length as an index of nigrostriatal dysfunction in mice. J. Neurosci. Methods113, 123–130 (2002).

28. Lin, A.H. et al. Global analysis of Smad2/3-dependent TGF-beta signaling in living micereveals prominent tissue-specific responses to injury. J. Immunol. 175, 547–554(2005).

29. Kim, Y.H., Choi, C.Y., Lee, S.J., Conti, M.A. & Kim, Y. Homeodomain-interacting proteinkinases, a novel family of co-repressors for homeodomain transcription factors. J. Biol.Chem. 273, 25875–25879 (1998).

30. Kim, Y.H., Choi, C.Y. & Kim, Y. Covalent modification of the homeodomain-interactingprotein kinase 2 (HIPK2) by the ubiquitin-like protein SUMO-1. Proc. Natl. Acad. Sci.USA 96, 12350–12355 (1999).

31. Kaartinen, V. et al. Abnormal lung development and cleft palate in mice lackingTGF-beta 3 indicates defects of epithelial-mesenchymal interaction. Nat. Genet. 11,415–421 (1995).

32. Proetzel, G. et al. Transforming growth factor-beta 3 is required for secondary palatefusion. Nat. Genet. 11, 409–414 (1995).

33. Brionne, T.C., Tesseur, I., Masliah, E. & Wyss-Coray, T. Loss of TGF-beta 1 leads toincreased neuronal cell death and microgliosis in mouse brain.Neuron40, 1133–1145(2003).

34. Bonyadi, M. et al. Mapping of a major genetic modifier of embryonic lethality in TGFbeta 1 knockout mice. Nat. Genet. 15, 207–211 (1997).

35. Shull, M.M. et al. Targeted disruption of the mouse transforming growth factor-beta 1gene results in multifocal inflammatory disease. Nature 359, 693–699 (1992).

36. Kele, J. et al. Neurogenin 2 is required for the development of ventral midbraindopaminergic neurons. Development 133, 495–505 (2006).

37. Andersson, E., Jensen, J.B., Parmar, M., Guillemot, F. & Bjorklund, A. Development ofthe mesencephalic dopaminergic neuron system is compromised in the absence ofneurogenin 2. Development 133, 507–516 (2006).

38. Lie, D.C. et al. The adult substantia nigra contains progenitor cells with neurogenicpotential. J. Neurosci. 22, 6639–6649 (2002).

39. Krieglstein, K. et al.Reduction of endogenous transforming growth factors beta preventsontogenetic neuron death. Nat. Neurosci. 3, 1085–1090 (2000).

40. Hofmann, T.G. et al. Regulation of p53 activity by its interaction with homeodomain-interacting protein kinase-2. Nat. Cell Biol. 4, 1–10 (2002).

41. D’Orazi, G. et al. Homeodomain-interacting protein kinase-2 phosphorylates p53 at Ser46 and mediates apoptosis. Nat. Cell Biol. 4, 11–19 (2002).

42. Gresko, E. et al. Autoregulatory control of the p53 response by caspase-mediatedprocessing of HIPK2. EMBO J. 25, 1883–1894 (2006).

43. Roussa, E. et al. Transforming growth factor beta is required for differentiation of mousemesencephalic progenitors into dopaminergic neurons in vitro and in vivo : ectopicinduction in dorsal mesencephalon. Stem Cells 24, 2120–2129 (2006).

44. Zhang, F., Endo, S., Cleary, L.J., Eskin, A. & Byrne, J.H. Role of transforming growthfactor-beta in long-term synaptic facilitation in Aplysia. Science 275, 1318–1320(1997).

45. Chin, J., Angers, A., Cleary, L.J., Eskin, A. & Byrne, J.H. Transforming growth factorbeta1 alters synapsin distribution and modulates synaptic depression in Aplysia.J. Neurosci. 22, RC220 (2002).

46. Rawson, J.M., Lee, M., Kennedy, E.L. & Selleck, S.B. Drosophila neuromuscularsynapse assembly and function require the TGF-beta type I receptor saxophone andthe transcription factor Mad. J. Neurobiol. 55, 134–150 (2003).

47. Aberle, H. et al. wishful thinking encodes a BMP type II receptor that regulates synapticgrowth in Drosophila. Neuron 33, 545–558 (2002).

48. McCabe, B.D. et al.Highwire regulates presynaptic BMP signaling essential for synapticgrowth. Neuron 41, 891–905 (2004).

49. Marques, G. et al. The Drosophila BMP type II receptor Wishful Thinking regulatesneuromuscular synapse morphology and function. Neuron 33, 529–543 (2002).


ART ICLES©

2007

Nat

ure

Pub

lishi

ng G

roup

ht

tp://

ww

w.n

atur

e.co

m/n

atur

eneu

rosc

ienc

e

CO R R I G E N D U M

NATURE NEUROSCIENCE

Corrigendum: Essential function of HIPK2 in TGFβ-dependent survival of midbrain dopamine neuronsJiasheng Zhang, Vanee Pho, Stephen J Bonasera, Jed Holzmann, Amy T Tang, Joanna Hellmuth, Siuwah Tang, Patricia H Janak, Laurence H Tecott & Eric J HuangNat. Neurosci. 10, 77–86 (2007); published online 10 December 2006; corrected after print 25 January 2007

In the version of this article initially published, the name of the fourth author was misspelled. The correct name is Jed Holtzman. The error has been corrected in the HTML and PDF versions of the article.

©20

07 N

atur

e P

ublis

hing

Gro

up

http

://w

ww

.nat

ure.

com

/nat

uren

euro

scie

nce

Essential function of HIPK2 in TGFβ-dependent survival of midbrain dopamine neurons

Documents

Transcript of Essential function of HIPK2 in TGFβ-dependent survival of midbrain dopamine neurons